Object detection device, object detection server, and object detection method

Abstract

Provided is a object detection device. A tag communication section (12) receives tag information transmitted from an information tag (11) attached to a person (P) to be detected. An attribute lookup section (15) looks up an attribute storage section (16) using ID included in the tag information to obtain attribute information such as the height of the person (P). A target detection section (14) specifies the position, posture and the like of the person (P) in an image obtained from an imaging section (13) using the attribute information.

Description

technical field [0001] The present invention relates to a technique for detecting the correct position, posture, etc. of a given detection object such as a person or an object from an image. Background technique [0002] Patent Document 1 describes a technique for controlling the orientation and zoom of a surveillance camera as a transmitter approaches. Specifically, a radio wave transmitter that transmits an ID code is attached to a person to be monitored, and an antenna that detects the proximity of the transmitter is installed in an area where entry is prohibited. When the antenna detects that the transmitter approaches, multiple surveillance cameras In this method, a surveillance camera near the position where the antenna can be photographed is automatically selected, and the image of the selected camera is displayed on the surveillance monitor. In addition, the ID code possessed by the transmitter is read through the antenna, and the height of the person corresponding ...

AI Technical Summary

Problems solved by technology

[0006] However, the previous technology is very difficult for such requirements

That is to say, in the above technology, as the transmitter approaches, only the camera that shoots the detection object can be selected, but detailed information such as where the detection object is and what posture it takes in the captured image cannot be obtained.

In addition, because the antenna is used to specify the position, there will be a relatively large error (several meters to more than ten meters) in the position information, that is, when there are other people near the detection object, it is difficult to identify two people in the image

[0007] In the past, several techniques for detecting objects from camera images using only image processing have been known, but most of them can only be used under very strictly limited conditions, and false detections may occur in the environment of people's daily life many, it is difficult to apply

The reasons for this are the limitations of the dynamic range of the camera itself, the existence of various objects and backgrounds other than the detection target, and changes in sunlight and lighting. kind of change etc.

[0008] The human visual mechanism can make use of a large amount of knowledge and rules acquired based on experience, and can perform correct detection even in a changing environment. However, it is still very difficult at this stage to make the device and the human obtain the same knowledge and rules. of

In addition, for such processing, a large amount of processing capacity and memory are required, which brings about problems such as an increase in processing time and cost.

Images